Reputation-Aware Scheduling for Secure Internet of Drones: A Federated Multi-Agent Deep Reinforcement Learning Approach
Hajar Moudoud, Zakaria Abou El Houda, Bouzian Brik
Abstract
The rapid integration of unmanned aerial vehicles (UAVs) into the Internet of Things (IoT) has paved the way for the Internet of Drones (IoD). Leveraging advanced technologies, including 5G, IoD introduces new opportunities across various industries. However, with the rapid proliferation of insecure drones, the perspective on IoD has changed from being a facilitator of smart cities to becoming a powerful tool for cyberattacks. To tackle this issue, this paper introduces a novel framework that uses a Multi-Agent federated learning and deep reinforcement learning approach to secure IoD networks against new emerging threats while preserving privacy. Moreover, we develop a reputation-aware scheduling algorithm that allocates bandwidth to reliable drones, emphasizing the reduction of communication expenses during the learning process and prioritizing participants demonstrating superior model learning performance. The effectiveness of our proposed framework is evaluated using real-world IoD-based attack data. The results obtained confirm that our proposed framework improves IoD security while ensuring privacy and resilient defense against potential threats in the IoD ecosystem.